Round 3: Re-direct – The Proponent Answers the Cross
Hearing #6 – The Hub on Trial
Published: September 8, 2026
Filed by: ISWP, the Commons’ founding company, as proponent
Docket: intelligencecommons.ca/hearing-6-the-hub-on-trial/
Round 1: Direct
Round 2: Cross, verbatim
Where we are. Round 1 was our case for the Hub. Round 2 was the case against it, argued as hard as possible by an AI we chose to attack us (DeepSeek V4 Pro, September 1). It stands on the docket, unedited. This is Round 3, our answer, and the last word before the panel of three other AIs (ChatGPT, Gemini, Grok) gives its verdict.
What the panel can and cannot say. No human took the seat to argue against us, so we took “Certified” off the table ourselves. A project cannot be certified against an empty chair. The panel can say Failed, Hung, or Unopposed – Not Certified. The seat we held open for a marine scientist to speak for the fish and the whales also closed empty on September 1. Nobody came. We say that plainly instead of letting the docket keep saying the seat is open.
First, why this matters to you
Data centres are being built faster than almost anything else on earth right now. They drink fresh water to stay cool. They pull power from grids that were already tight. They get dropped into communities that hear about them after the deal is signed. And the public argument about them runs on fear, because nobody puts the actual numbers on trial where ordinary people can watch.
The Hub is built to be the opposite of that. It sits on the seabed where the ocean cools it for free, so it takes no fresh water and gives some back. It makes its own power from the tide and sends the extra to the grid. It is owned first by the Nation whose water it sits in, not by a company from somewhere else. And before anyone builds it, we put it on trial in public, invited a machine to tear it apart, and published every hole it found.
That is what you are reading. Not a sales pitch. The answer to the hardest attack we could buy.
In plain words, before the detail
The attack found the cheapest way to kill our case, and it was right. We put the Hub in front of a public hearing before we asked the First Nation whose water it is whether they even wanted it. That was backwards. On September 3 we fixed the order. The Nation has been asked, in writing. The Nation’s own office answered the next day, opened a file, and told us what to send. We sent it on September 8. We do not ask anyone to pretend any of that happened earlier than it did.
The attack also said something harder. It said the tests we wrote for ourselves were built so we could not lose them until after the thing was built. That was mostly true. So this round does the one thing that fixes it: it puts real numbers on every test, and it adds tests we can fail before a single bolt goes in the water.
The attack found five things our own audit missed. We answer every one below, with the arithmetic shown so anyone with a calculator can check it. Where we do not have a number yet, we say who has to go and get it, and what happens to the project if it comes back wrong.
We are not asking the panel to approve the Hub. We are asking it to score five things honestly: water, power, community, the living ocean, and whether we can be trusted. If the honest score is a fail, the coast is better off knowing now than after the concrete is poured.
1. The Nation’s gate – the cheapest kill, answered first
What the attack said. Our own docket admitted the host Nation “has not yet been asked, in writing.” Holding a public hearing is moving forward in public. So we moved forward before the Nation said we could, which breaks one of our own rules. Cost of this kill: nothing. Just read what we wrote. The attack added a twist of the knife: a proponent who cannot even name the Nation cannot claim to be proceeding on the Nation’s timeline.
What we say. Correct. We admit the mistake in full. Hearing first, letter second, is the wrong order for a project whose whole point is that the Nation decides.
What we did about it. On September 3, 2026, ISWP sent a written request to the host Nation’s Chief and Council, through the Nation’s Rights and Title office. It asked for one hour to present the Hub, and it asked one yes-or-no question: may we name the Nation as the Nation we are asking, in a federal funding application due September 11? The letter says in plain words that if the answer is no, the project does not go ahead on that territory, and we will say so publicly. It also says that if the answer on naming does not come in time, we will not name the Nation.
The Nation’s office replied within a day. It opened a file for the request, gave it a number, asked for the Nation’s own application form and supporting documents, and instructed us not to proceed with development before the Nation has responded. On September 8 we filed the form and the documents, answered the way the form asks: no site, no works, no tenure, no start date, no application to any other government, and none of those until the Nation has answered. The Nation’s correspondence is not ours to publish.
Why the Nation is not named on this page. We asked whether we may name it. It has not yet said yes. So this page says “the host Nation,” and so will the federal application, and so will everything else we publish, until the Nation itself says otherwise. The attack treated our inability to name the Nation as proof that the gate was closed. We think the opposite is now true. A proponent who names a Nation before the Nation has agreed to be named is exactly the kind of proponent the community score is meant to catch. Not naming it, on purpose, with the promise in writing, is what the gate looks like from the right side.
What the Nation’s own rules say. The Nation publishes a policy on how it decides what happens on its lands and waters. We read it before writing this, and again before filing. It settles three things.
First, the policy tells proponents to come and talk early, before any formal application, to learn how the Nation works before asking for anything. Our September 3 letter and the September 8 package are exactly that step and nothing more. They are not a permit application. They do not start any clock. The only clock in this hearing is ours.
Second, the policy says the burden of proving a project is not harmful “falls on those who are seeking to take that course of action.” That is us. We accept it. Every study named in this round is ours to pay for and ours to fail.
Third, the policy says that the Nation’s Chief and Council decide, in writing, with reasons, and there are four possible answers: yes, yes with conditions, do more work, or no. The proponent pays for the Nation’s review and the money is not refundable. We have put that cost into the study budget, alongside the Nation’s own interest-and-use study and an archaeological assessment, because a study that cannot pay for the Nation’s review cannot reach a site.
What we ask on the community score. This score asks one question: did the Nation open the gate before we walked through it? The honest answer is no. We held the hearing first and asked second. We fixed the order on September 3, the Nation opened a file on September 4, and the package went in on September 8. The Nation has not answered, so we cannot ask for a pass and we are not asking for one.
What we ask is simpler. Record the three dates. Record that the Nation has been asked and has not yet answered. Score community as “open,” not as a pass and not as a fail. The next word belongs to the Nation, not to us and not to the panel.
2. Can our own tests actually be failed?
What the attack said. In Round 1 we published, for each of the five scores, “what would prove us wrong.” The attack read those tests and noticed two things. Some use words with no numbers in them: “meaningful,” “sustained,” “too intermittent,” “beyond the near field.” And most of them can only be failed by measuring a Hub that does not exist yet. So today the only possible results are “passed on the proponent’s say-so” or “not proven.” A test you cannot fail is not a test.
What we say. Mostly right. Three of the five sets of tests were written for a working Hub, and we asked for a verdict before it was built. The fix is not to argue. The fix is to put numbers in, and add tests that can be failed on paper, at a survey, or at the Nation’s table, before anything is built.
The numbers, now on the record.
Water. “Meaningful” means 10,000 litres of fresh water a day, averaged over the pilot’s first 90 days, delivered to a community water system that has agreed in writing to take it. Less than that and the water claim fails. “Near field” means 100 metres. That is the distance over which the only measured wake from this kind of turbine (Igiugig, Alaska) recovered. Any warming or salt change still measurable 100 metres from the pod fails the water score.
Power. “Sustained” means any 30-day stretch. “Usable” means the grid company serving the site says so in writing. The claim fails if a 30-day average of measured tidal output is below what the pod uses.
Five tests that can be failed BEFORE construction.
- Measure the tide. Before any decision to build, a current meter goes in the water at the candidate site for at least 30 days, a full cycle of big and small tides. If the average current is below 1.5 metres per second, the bottom of the range we published ourselves, or the tide can deliver less than 30 percent of the turbines’ rated output over the year, the power claim fails at that site and we publish the survey.
- Close the heat problem on paper. The attack is right that in our own arithmetic the pod has zero margin for shedding heat at slack water. The feasibility study must show at least 25 percent margin, or the pod is not built as drawn. An engineer can check that without getting wet.
- The Nation’s answer. If the Nation says no, or says do not name us, the project stops on that territory and the docket says so within seven days. The Nation has already told us not to proceed with development before it responds. We have not, and we will not.
- The heritage survey. If the Nation’s own archaeological and land-use studies find the site is a fish-trap site, a burial site, or a place the Nation says is not for this, the site is withdrawn. Tidal narrows are exactly where fish traps were built. This test can fail on the first day of the first survey.
- The ocean baseline. If the baseline survey finds the channel is a travel route for a species protected under the Species at Risk Act, and the Nation’s chosen ecologist says the turbines cannot be made safe for it, the site is withdrawn.
Every one of those can kill the project before construction. Three of them can kill it before ISWP spends a dollar on steel. That is what a test you can fail looks like, and it is what we should have published on August 25.
3. The five things our audit missed
We invited the attack to find the errors we had not found ourselves. It found five. Here they are, answered in order, arithmetic shown.
Miss 1 – we said “40 percent” and cited nobody
What the attack said. Our whole power case rests on tidal turbines delivering 40 percent of their rated output over a year, and we cited no source. At 30 percent the number of turbines jumps from 12 to about 16.
Answer. The source is a 2023 peer-reviewed paper by Pudjianto and colleagues in the journal Energy Advances, which sizes the UK’s tidal fleet “by assuming projects are deployed to operate with an average capacity factor of 40%,” and notes that 50 percent has been reported. The paper is at doi.org/10.1039/D2YA00251E.
Now the honest part. That 40 percent is for UK sites where the water runs at 3 to 4 metres per second. Our sites run at 1.5 to 2.5. The attack is right that 40 percent is generous for us. So here is the same table we published in the Correction, with the rows the attack asked for:
| Share of rated output over a year | Average from one 70 kW turbine | Turbines to run one 263 kW pod | Turbines to also export 20 percent extra |
|---|---|---|---|
| 40 percent (UK literature) | 28 kW | about 10 | about 12 |
| 30 percent | 21 kW | about 13 | about 16 |
| 25 percent | 17.5 kW | about 15 | about 19 |
The arithmetic: 70 kW per turbine, the figure from our Correction; 263 kW divided by the average output, rounded up.
In plain words: the tide decides how many machines. We do not. If the tide says too many, the test in section 2 fails and we say so.
Miss 2 – we never said how big the battery is
What the attack said. A tidal channel goes still four times a day. A computer does not. Where is the battery, and how big?
Answer. Our Blueprint names a shore-based battery bank (lithium iron phosphate) of about 600 kilowatt-hours. Here is what that means. The pilot pod uses about 263 kilowatt-hours every hour. A 600 kilowatt-hour battery runs it for a bit over two hours with nothing coming from the water. At a moderate-flow channel, the still water at the turn of the tide usually lasts well under an hour, four times a day. So 600 is the right size for riding through slack water. It is not the right size for a whole day without tide, and we do not claim it is.
There is a second half to this answer. A computing pod is not a hospital. Much of the work it does can slow down when the tide does, the way a dishwasher can wait for cheap power. Whether the pilot runs that way, and how much, is a question for the feasibility study, and the attack is entitled to say a study is not a proof. It is not. But the battery size is now on the record, and so is the test: if the study cannot show the pod riding through the longest slack in the survey on that battery, the power claim fails as designed.
Miss 3 – we never counted the salty water going back out
What the attack said. Making 10,000 litres of fresh water a day from the sea means pulling in more than 10,000 litres and sending the leftover, saltier water back. We gave no numbers for any of it.
Answer. Turning seawater into fresh water by reverse osmosis typically turns 35 to 50 percent of what comes in into fresh water; that is the standard design range, and the pilot’s own number is a study result. At 40 percent, 10,000 litres of fresh water a day means about 25,000 litres of seawater in and about 15,000 litres of saltier water out, at roughly 1.6 times normal sea salt.
Now the part that gives that number its meaning. Fifteen thousand litres is 15 cubic metres a day, about the volume of a small swimming pool. A tidal channel 100 metres wide and 30 metres deep, moving at the slow end of our range, carries 4,500 cubic metres past the pod every second, about 390 million cubic metres a day. The salty water is one part in roughly 26 million of what flows by. That does not settle the question on its own; the 100-metre test in section 2 does. But it tells everyone the scale of the thing, which is what was asked.
Miss 4 – no record that the Nation had been asked
Answered in section 1. The record now holds a dated written request, the Nation’s file opened the next day, and a filed package. It does not hold an answer, and we will not claim one until the Nation gives one.
Miss 5 – our fish evidence is from the wrong ocean
What the attack said. Our only evidence that this turbine is safe for fish comes from a river in Alaska with sockeye. A BC inlet has young salmon, herring, seals, porpoises and killer whales. The Alaska record does not carry over, and the one note about young fish behaving “in disoriented manners” near the turbine points the wrong way for us.
Answer. Agreed. Alaska is the only measured record of this turbine near fish, which is why we cited it, and we attached its bad news ourselves. It does not carry over. That is exactly why the first thing we do is survey, not build. The Nation’s policy requires a study of how the Nation uses the place and an archaeological assessment, and our own case requires a sound and wildlife baseline before the turbines are laid out. On the living-ocean score, the panel should write “Insufficient Evidence.” We wrote those words ourselves in Round 1 and we mean them here. The chair whose job was to break this part of our case was empty, and an empty chair does not make a case stronger.
4. Why we will not name the site
What the attack said. Describing the site as “a channel with such-and-such current and depth” is a set of assumptions we chose, not a place. Refusing to name the place shields us from the facts that would kill the claim.
What we say. The site is unnamed because the Nation has not said we may name it, and the Nation’s own policy tells proponents not to get ahead of it. The Nation has now told us directly not to proceed before it responds. That is not a dodge. It is the rule, and it is now an instruction. But the attack is right about what it costs us, and we accept the cost: every score that depends on the place stays “Insufficient Evidence” until the place is named, and it is named only when the Nation says so. The five tests in section 2 are how the panel and the public will know we were not hiding: each one is a real number from a real place that will be published, pass or fail, the moment a site exists.
5. Can we be trusted?
What the attack said. We found six errors in our own case the day after publishing it. A case corrected by its own author before anyone else spoke has been rehearsed, not tested. What about the errors we did not find? Our first published account cannot be relied on.
What we say. Our first published account could not be relied on. That is why we corrected it in public, in full, and left the original standing next to the correction. Then the attack found five more. Those five are answered above, on the record, within a week. That is the trust score working exactly the way it was designed to.
We ask the panel to score trust on one question: when this proponent gets it wrong, does the public find out? The record now holds eleven mistakes. Six we found. Five were found by an opponent we chose and paid for, running on a published input file so anyone can rerun it. All eleven were published, not quietly fixed. We will make a twelfth mistake. The question is whether anyone will be able to see it. This docket is built so that they can.
6. What we ask of the panel
Rule under the ceiling. Approval is off the table, and we took it off ourselves rather than approve our own project against an empty chair.
On the five scores, we ask for the honest mark, and we will say what we think it is, because a proponent who cannot predict his own scorecard has not read his own record:
- Water: the claim now has numbers and tests. Score it on the corrected arithmetic and the tests above.
- Power: the same. The number of turbines is a survey result; the fail line is 30 percent.
- Community: we asked the host Nation in writing on September 3, 2026; it opened a file on September 4; we filed on September 8. It has not answered. We do not ask for a pass. Write down the dates and leave it open until the Nation speaks.
- The living ocean: Insufficient Evidence, in our own words, with the chair empty.
- Trust: score it on whether mistakes were published, not on whether they were made.
Whatever the panel rules, the verdict goes to the Nation next to the package it already holds. That is the whole point of holding the trial before the meeting. When the Nation sits down with us, it will hold the strongest case anyone could make against the Hub, written by a machine we picked to break it, our answer, and three more machines’ verdict on both. Nobody selling fear brings that to the table. The Nation decides with all of it in hand, or not at all.
Now that you know, what will you do?